Change
49ca537
49ca5372c742edcd64b339ab2cdc3357d013c87d · commit on GitHub
pytorch-blog-feed: changed (347092 bytes, HTTP 200)
raw/pytorch-blog-feed/response.xml modified
- Source
- pytorch-blog-feed
- Lines added
- +44
- Lines removed
- -185
- Stored bytes at this commit
- 347,092
- Timestamp
- origin
- Raw artifact at this commit
- raw/pytorch-blog-feed/response.xml
Recorded headers
| observed_at | 2026-09-18T04:44:07.546Z |
|---|---|
| origin_date | 2026-09-18T04:44:03.000Z |
| status | 200 |
| final URL | https://pytorch.org/blog/feed/ |
| etag | "f78f9c94629e75c1ecd6d4824bf464d3" |
| last-modified | Fri, 18 Sep 2026 00:27:15 GMT |
| date | Fri, 18 Sep 2026 04:44:07 GMT |
| age | 4 |
| cache-control | public, max-age=60, s-maxage=43200, stale-while-revalidate=86400, stale-if-error=604800 |
| cf-cache-status | null |
| content-encoding | null |
| content-length | 347092 |
@
@@ -12,7 +12,7 @@ <atom:link href="https://pytorch.org/blog/feed/" rel="self" type="application/rss+xml" /> <link>https://pytorch.org</link> <description></description>-
<lastBuildDate>Wed, 16 Sep 2026 22:48:18 +0000</lastBuildDate>+
<lastBuildDate>Fri, 18 Sep 2026 00:27:15 +0000</lastBuildDate> <language>en-US</language> <sy:updatePeriod> hourly </sy:updatePeriod>@
@@ -28,6 +28,48 @@ <height>32</height></image> <item>+
<title>PyTorch Day Japan 2026 Comes to Tokyo on December 10</title>+
<link>https://pytorch.org/blog/pytorch-day-japan-2026-comes-to-tokyo/</link>+
+
<dc:creator><![CDATA[PyTorch Foundation]]></dc:creator>+
<pubDate>Fri, 18 Sep 2026 00:25:32 +0000</pubDate>+
<category><![CDATA[Announcements]]></category>+
<category><![CDATA[Blog]]></category>+
<guid isPermaLink="false">https://pytorch.org/?p=168823</guid>+
+
<description><![CDATA[PyTorch Day Japan 2026 will bring the open source AI community together in Tokyo on December 10 for a full day of technical talks and interactive discussions designed to foster...]]></description>+
<content:encoded><![CDATA[<p><a href="https://events.linuxfoundation.org/pytorch-day-japan/">PyTorch Day Japan 2026</a> will bring the open source AI community together in Tokyo on December 10 for a full day of technical talks and interactive discussions designed to foster knowledge exchan…+
<p>Hosted by PyTorch Foundation, Hugging Face, IBM, and Mitsubishi Electric, the event will bring together PyTorch enthusiasts, machine learning engineers, AI researchers, and industry professionals working across open source AI.</p>+
<p>The program will explore PyTorch and PyTorch Foundation hosted projects, including vLLM, DeepSpeed, Ray, Helion, and Safetensors, alongside broader topics in AI and machine learning such as training, inference, responsible AI, physical and edge AI, and open model development.</p>+
<p><a href="https://events.linuxfoundation.org/pytorch-day-japan/program/cfp/">The call for proposals is now open</a>, and <a href="https://events.linuxfoundation.org/pytorch-day-japan/register/">registration is available with discounted pricing</a> through November 11.</p>+
<h2>Submit a Session Proposal</h2>+
<p>PyTorch Day Japan 2026 is accepting proposals for session presentations and lightning talks.</p>+
<p>Suggested topics include:</p>+
<p><strong>Sovereign AI and local open models:</strong> Strategies and architectures for building and running open-weights AI on local infrastructure, including open model adaptation, post-training, local inference, and privacy-first deployments using PyTorch.</p>+
<p><strong>Physical AI and edge AI:</strong> Experiences bringing PyTorch models to physical hardware, robotics, and edge systems, including real-time on-device inference, vision-language-action models, hardware acceleration, and optimization for resource-constrained environments.</p>+
<p><strong>PyTorch ecosystem:</strong> Developments across the PyTorch library and developer toolchain, including domain libraries such as TorchVision, TorchAudio, TorchRL, and PyTorch Distributed, as well as developer tooling, PyTorch 2.x and <code>torch.compile</code>, production pipelines, and co…+
<p>The CFP closes <strong>Sunday, September 27 at 11:59 PM JST</strong>.</p>+
<p>Key dates:</p>+
<ul>+
<li>CFP deadline: Sunday, September 27 at 11:59 PM JST</li>+
<li>CFP notifications: Tuesday, October 13</li>+
<li>Schedule announcement: Wednesday, October 14</li>+
<li>Presentation slides due: Wednesday, December 9</li>+
<li>PyTorch Day Japan: Thursday, December 10</li>+
</ul>+
<p><a href="https://events.linuxfoundation.org/pytorch-day-japan/program/cfp/">Submit a proposal</a></p>+
<h2>Register for PyTorch Day Japan</h2>+
<p>Registration is also open for PyTorch Day Japan 2026.</p>+
<p>Registration is <strong>¥5,000 through November 11 at 11:59 PM JST</strong>, representing ¥3,000 in savings. Discounted academic pricing is also available for students and faculty members.</p>+
<p><a href="https://events.linuxfoundation.org/pytorch-day-japan/register/">Register for PyTorch Day Japan</a></p>+
<p>PyTorch Day Japan 2026 takes place <strong>Thursday, December 10 in Tokyo, Japan</strong>.</p>+
<p>For full event details, visit the <a href="https://events.linuxfoundation.org/pytorch-day-japan/">PyTorch Day Japan 2026 website</a>.</p>+
]]></content:encoded>+
+
+
+
</item>+
<item> <title>Open Research, Tooling & Optimization at PyTorch Conference North America 2026</title> <link>https://pytorch.org/blog/open-research-tooling-optimization-at-pytorch-conference-north-america-2026/</link> @
@@ -1822,191 +1864,8 @@ for n in [1024, 1 << 16, 1 << 20, 1 << 24]: -
</item>-
<item>-
<title>Agentic AI and Next-Gen Intelligence Sessions at PyTorch Conference North America 2026</title>-
<link>https://pytorch.org/blog/agentic-ai-and-next-gen-intelligence-sessions-at-pytorch-conference-north-america-2026/</link>-
-
<dc:creator><![CDATA[PyTorch Foundation]]></dc:creator>-
<pubDate>Wed, 02 Sep 2026 16:50:44 +0000</pubDate>-
<category><![CDATA[Announcements]]></category>-
<category><![CDATA[Blog]]></category>-
<category><![CDATA[Agentic AI]]></category>-
<category><![CDATA[Next-gen Intelligence]]></category>-
<category><![CDATA[PyTorch Conference]]></category>-
<category><![CDATA[PyTorch Conference North America]]></category>-
<guid isPermaLink="false">https://pytorch.org/?p=161495</guid>-
-
<description><![CDATA[TL;DR PyTorch Conference North America 2026 features Agentic AI and Next-Gen Intelligence across sessions on training agents, serving agents in production, agents that build PyTorch, and PyTorch in the physical...]]></description>-
<content:encoded><![CDATA[<h3><b>TL;DR</b></h3>-
<p><span style="font-weight: 400;">PyTorch Conference North America 2026 features Agentic AI and Next-Gen Intelligence across sessions on training agents, serving agents in production, agents that build PyTorch, and PyTorch in the physical world.</span></p>-
<h2><b>Agentic AI and Next-gen Intelligence at PyTorch Con NA</b></h2>-
<p><span style="font-weight: 400;">When you look at the schedule for PyTorch Conference North America 2026, one thing jumps out: agents are featured everywhere. They write kernels, triage CI, migrate workloads between chips, drive robots, and answer the phone. The interesting questions have shifted …-
<p><span style="font-weight: 400;">Here is a guided tour of the agentic AI and next-generation intelligence content across the two days of PyTorch Conference North America, along with why it is worth being in the room.</span></p>-
<p><a href="https://hubs.ly/Q04tDx8f0"><span style="font-weight: 400;">View the full conference schedule</span></a><a href="https://www.google.com/search?q=https://hubs.ly/Q04tDx8f0"><span style="font-weight: 400;"><br />-
</span></a><a href="https://hubs.ly/Q04tDw_W0"><span style="font-weight: 400;">Register for PyTorch Conference North America 2026</span></a></p>-
<h2><b>It starts on the keynote stage</b></h2>-
<p><span style="font-weight: 400;">Three keynotes frame the whole conversation.</span></p>-
<h3><b>Beyond Brute Force: The Era of Adaptive Intelligence</b></h3>-
<p><strong>Sara Hooker, Adaption</strong><br />-
<i>October 20, 09:40am | Grand Ballroom</i></p>-
<p><span style="font-weight: 400;">The next unlock isn’t scale, it’s architecture: systems that keep learning after deployment, closing the gap between a model’s frozen training distribution and the world it actually operates in. Sara’s talk digs into continual, gradient-free learn…-
<h3><b>Workload Fungibility in the Age of Agents</b></h3>-
<p><strong>Bill Jia, Google Cloud</strong><br />-
<i>October 21, 09:15am | Grand Ballroom</i></p>-
<p><span style="font-weight: 400;">The other half of the story addresses agents as developers. Alongside the deep dive on TorchTPU going open source, Jia demos long-horizon agentic workflows that migrate complex model workloads from GPUs to TPUs. The workflows keep going, hill-climbing on quantizati…-
<h3><b>Agentic AI Foundation Keynote</b></h3>-
<p><strong>Mazin Gilbert, Agentic AI Foundation</strong><br />-
<i><span style="font-weight: 400;">October 21, 09:10 | Grand Ballroom</span></i></p>-
<p><span style="font-weight: 400;">Mazin Gilbert is the Executive Director of the Agentic AI Foundation at the Linux Foundation and has over 25 years of experience pioneering open source platforms, authoring 100+ research papers, and holding 260+ U.S. patents. In his keynote, he will provide critica…-
<h2><b>Training agents: RL becomes a requirement </b></h2>-
<p><span style="font-weight: 400;">The single densest cluster of agentic content is in post-training. Multi-turn, tool-using, long-horizon RL has moved from research curiosity to production requirement.</span></p>-
<h3><b>Agentic RL Training in PyTorch</b></h3>-
<p><strong>Yichuan Wang, Shuhua Yu, Meta</strong><br />-
<i><span style="font-weight: 400;">October 20, 16:20 | 210AE</span></i></p>-
<p><span style="font-weight: 400;">An end-to-end overview of the reinforcement learning training loop, spanning rollout infrastructure, trainer-serving communication, environment abstractions, and sandbox execution, alongside task scheduling and the trade-offs between on-policy and off-policy method…-
<h3><b>Open Source Reinforcement Learning with Agent Harnesses</b></h3>-
<p><strong>Ben Burtenshaw, Hugging Face</strong><br />-
<i><span style="font-weight: 400;">October 20, 17:30 | 210AE</span></i></p>-
<p><span style="font-weight: 400;">OpenEnv is the interoperability layer for publishing and running RL environments, co-owned by Hugging Face, Meta, Unsloth, Prime Intellect, Modal, NVIDIA, Mercor, and others. Frontier labs train models inside their own harness, while the open ecosystem vendors mode…-
<h3><b>Train the Agent, Not Just the Model</b></h3>-
<p><strong>Sergio Paniego Blanco, Hugging Face</strong><br />-
<i><span style="font-weight: 400;">October 21, 12:20 | LL21DEF</span></i></p>-
<p><span style="font-weight: 400;">Explore practical SFT and GRPO techniques in agentic environments, progressing to harness-driven training where the harness manages its own inner loop directly inside the environment.</span></p>-
<h3><b>Torchtitan RL: A Unified and Extensible Training Framework for Agentic Tasks</b></h3>-
<p><strong>Felipe Mello, Jiani Wang, Meta</strong><br />-
<i><span style="font-weight: 400;">October 21, 16:20 | LL21DEF</span></i></p>-
<p><span style="font-weight: 400;">In Felipe and Jiani’s session, attendees learn how to use a single model definition for both training and generation with an on-policy, bitwise-reproducible mode that keeps system artifacts from getting in the way of your reward design.</span></p>-
<h3><b>Miles: Enterprise-facing Agentic RL Framework</b></h3>-
<p><strong>Mao Cheng, RadixArk</strong><br />-
<i><span style="font-weight: 400;">October 21, 11:45 | LL21DEF</span></i></p>-
<p><span style="font-weight: 400;">Learn how to scale post-training using unified low-precision training, stable Mixture-of-Experts (MoE) reinforcement learning, and accelerated speculative rollouts.</span></p>-
<h3><b>When Rollout and Training Disagree</b></h3>-
<p><strong>Neiwen Ling (ByteDance), Tianle Zhong (University of Virginia)</strong><br />-
<i><span style="font-weight: 400;">October 21, 16:55 | LL21DEF</span></i></p>-
<p><span style="font-weight: 400;">A sharp, specific lightning talk on training-inference mismatch: why small token-level numerical disagreements between the rollout engine and the training path are not benign, and how they can quietly reshape your PPO/GRPO objective.</span></p>-
<h2><b>Serving agents in production</b></h2>-
<p><span style="font-weight: 400;">Agentic workloads break the assumptions inference stacks were built on. Sessions may sit idle for hours before a follow-up arrives. Context is long, multi-turn, and tool-laden. CPU work such as orchestration, tool execution, and scheduling stops being a rounding er…-
<h3><b>Making Enterprise Agentic Inference Production-Ready with PyTorch and vLLM</b></h3>-
<p><strong>Joseph Groenenboom, Tyler Michael Smith, Red Hat</strong><br />-
<i><span style="font-weight: 400;">October 20, 15:25 | LL21ABC</span></i></p>-
<p><span style="font-weight: 400;">Joseph shares what enterprise readiness actually entails, from build infrastructure up through tool calling and long-context multi-turn chat, with practical insights from the engineers themselves. </span></p>-
<h3><b>Sponsored: PyTorch for Agentic AI: Scaling Heterogeneous Systems from CPU to XPU</b></h3>-
<p><strong>Eikan Wang, Huma Abidi, Intel</strong><br />-
<i><span style="font-weight: 400;">October 20, 17:30 | LL21DEF</span></i></p>-
<p><span style="font-weight: 400;">To accommodate system-intensive and heterogeneous agentic AI workloads, Intel employs an upstream-first strategy across Intel Xeon processors, Intel GPUs, and its open software stack to deliver standard PyTorch workflows and deep performance optimizations.</span></…-
<h3><b>Native Tiered KV Cache Offloading in vLLM</b></h3>-
<p><strong>Or Ozeri, IBM</strong><br />-
<i><span style="font-weight: 400;">October 20, 14:50 | LL20CD</span></i></p>-
<p><span style="font-weight: 400;">To address LLM scaling challenges with long-lived agentic sessions, vLLM introduces an upstream, dependency-free tiered KV cache offloading framework that routes transfers through CPU memory as a universal transport hub to minimize GPU overhead and ensure hardware-…-
<h3><b>LMCache: a cluster-wide open source solution for LLM prompt caching</b></h3>-
<p><strong>Kuntai Du, Tensormesh</strong><br />-
<i><span style="font-weight: 400;">October 20, 15:40 | LL20CD</span></i></p>-
<p><span style="font-weight: 400;">LMCache offers a popular, open source prompt caching solution featuring extensive support across major inference engines and storage backends, alongside Kubernetes deployment guidance and underlying research insights. </span></p>-
<h2><b>Agents that build PyTorch</b></h2>-
<p><span style="font-weight: 400;">This is the most self-referential thread on the schedule, and one of the most practical.</span></p>-
<h3><b>Contributing to PyTorch with AI agents (Birds of a Feather)</b></h3>-
<p><strong>Edward Yang, Meta</strong><br />-
<i><span style="font-weight: 400;">October 20, 10:35 | Community Expo</span></i></p>-
<p><span style="font-weight: 400;">How should agents be used to contribute productively? How do you get your PR reviewed? Come, discuss and share your perspectives in person.</span></p>-
<h3><b>Fighting Agents with Agents: Bringing Claude to PyTorch CI, triage, and PR review</b></h3>-
<p><strong>Driss Guessous, Meta</strong><br />-
<i><span style="font-weight: 400;">October 20, 12:35 | LL21DEF</span></i></p>-
<p><span style="font-weight: 400;">Maintainers are already reviewing an ever-increasing number of agent-written PRs. This is the story of giving maintainers agent-shaped infrastructure for an agent-shaped world: issue triage, PR review skills, autorevert investigation, and the adoption curve after l…-
<h3><b>Shipping PyTorch and Its Ecosystem: A Modern Release Story</b></h3>-
<p><strong>Andrey Talman, Meta</strong><br />-
<i><span style="font-weight: 400;">October 20, 11:45 | LL21DEF</span></i></p>-
<p><span style="font-weight: 400;">PyTorch functions as a unified release train that integrates ecosystem projects like Triton and vLLM to deliver faster, highly validated, and predictable software builds through upstream continuous integration and agent-assisted triage. Join this session for an hon…-
<h2><b>Kernel and Performance Agentic Search</b></h2>-
<p><span style="font-weight: 400;">Agentic search is producing real performance numbers across kernel work.</span></p>-
<h3><b>KernelAgent: Hardware-Guided GPU Kernel Optimization via Multi-Agent Orchestration</b><span style="font-weight: 400;"> </span></h3>-
<p><strong>Kaiming Cheng, Laura Wang, Meta</strong><br />-
<span style="font-weight: 400;">October 20, </span><i><span style="font-weight: 400;">14:15 | LL21DEF</span></i></p>-
<p><span style="font-weight: 400;">Building on its 100% correctness benchmark across KernelBench tasks, the updated open source KernelAgent integrates GPU hardware-performance signals into a closed-loop multi-agent workflow to optimize Triton kernels, delivering a 1.56x average speedup over default …-
<h3><b>Smarter Autotuning for Kernels: From Bayesian Optimization to LLM-Guided Search in Helion DSL</b></h3>-
<p><strong>Jongsok Choi, Ethan Che, Meta</strong><br />-
<span style="font-weight: 400;">October 20, </span><i><span style="font-weight: 400;">14:15 | 210BF</span></i></p>-
<p><span style="font-weight: 400;">PyTorch’s Helion DSL transforms compile-time kernel autotuning by combining Likelihood-Free Bayesian Optimization (LFBO) and LLM-guided search into an LLM-seeded hybrid approach that delivers up to 10X faster tuning times and improved performance on NVIDIA H1…-
<h3><b>Helion: CuteDSL and TPU Backends for Heterogeneous Hardware, and Why It Suits Agents</b></h3>-
<p><strong>Oguz Ulgen, Dunfan Lu, Jason Ansel, Meta</strong><br />-
<span style="font-weight: 400;">October 20,</span><i><span style="font-weight: 400;"> 11:45 | 210BF</span></i></p>-
<p><span style="font-weight: 400;">Helion is a high-level Python DSL for writing performant, portable ML kernels across NVIDIA GPUs and TPUs using CuteDSL and Pallas backends, while also leveraging LLM agents and built-in autotuning to simplify code generation. Join this session to learn why CuteDSL…-
<h3><b>From Weeks to Overnight: Autonomous Day-0 Kernel Bring-Up with Agent Pipelines</b></h3>-
<p><strong>Xiaogang Gu, Qun Yang, Intel</strong><br />-
<i><span style="font-weight: 400;">October 20, 17:30 | 210BF</span></i><span style="font-weight: 400;">. </span></p>-
<p><span style="font-weight: 400;">This session examines how to replace repetitive manual kernel optimization with an autonomous system that uses context-isolated specialized agents across a deterministic profiling-to-benchmarking workflow with long-term memory. This reduces vLLM kernel bring-up tim…-
<h3><b>Primus Tuning: Hybrid Projection and Agentic Search for Distributed Training</b></h3>-
<p><strong>Anshu Raina, Peyman Razaghi, AMD</strong><br />-
<span style="font-weight: 400;">October 21,</span><i><span style="font-weight: 400;"> 16:20 | 210AE</span></i></p>-
<p><span style="font-weight: 400;">The Primus Tuning Agent automates distributed LLM training configurations by using a single-node hybrid projection engine to analytically reconstruct cross-node performance within ~10% accuracy, driving an LLM-guided search that boosted Mixtral 8x22B throughput by …-
<h3><b>Sponsored: Cloud TPU Agent Suite: Autonomous Multi-Agent Swarms for PyTorch<br />-
</b></h3>-
<p><strong>Sandeep Pokkunuri, Chris Jones, Google</strong><br />-
<i><span style="font-weight: 400;">October 20, 11:45 | 210AE</span></i></p>-
<p><span style="font-weight: 400;">Cloud TPU Nexus streamlines PyTorch model migration from GPUs to TPUs by embedding an autonomous multi-agent intelligence platform directly into developer IDE tools, achieving over 70% of hand-tuned performance in under 24 hours without manual kernel tuning. </span…-
<h3><b>TorchInsights: Zero-GPU Memory & Runtime Estimation</b></h3>-
<p><strong>Sanket Jayant Purandare, Aditya Venkataraman, Meta</strong><br />-
<i><span style="font-weight: 400;">October 20, 17:45 | LL21ABC</span></i><span style="font-weight: 400;">. </span></p>-
<p><span style="font-weight: 400;">Published at ICML 2025 as TorchSim, the open source TorchInsights tool simulates multi-stream GPU execution via zero-GPU fake tensor execution to accurately break down peak memory and profile distributed runtime configurations before launching multi-node jobs.</spa…-
<h2><b>Agents in the physical world</b></h2>-
<p><span style="font-weight: 400;">Next-gen intelligence isn’t only text. A significant chunk of the schedule focuses on models that see, hear, and move.</span></p>-
<h3><b>From Pixels to Physical Motion: Building World Action Models with PyTorch & NVIDIA Cosmos</b></h3>-
<p><strong>Susie Xia, Ruijie Zheng, George Kurian, NVIDIA</strong><br />-
<i><span style="font-weight: 400;">October 21, 14:15 | 210AE</span></i></p>-
<p><span style="font-weight: 400;">NVIDIA Cosmos and GR00T serve as interconnected case studies for building physical AI in PyTorch, showing how to translate multimodal world understanding into executable robot motion while navigating end-to-end training and latency-sensitive deployment challenges. …-
<h3><b>Deploying Robot Policies Across Many Targets with ExecuTorch</b></h3>-
<p><strong>Jacob Szwejbka, Meta</strong><br />-
<i><span style="font-weight: 400;">October 21, 15:25 | 210AE</span></i></p>-
<p><span style="font-weight: 400;">ExecuTorch provides a production deployment layer that translates PyTorch robot policies from frameworks like LeRobot or OpenVLA into portable runtime artifacts across NVIDIA, Intel, ARM, and microcontroller hardware without requiring per-target stack rebuilds. </s…-
<h3><b>Multimodal Dataloaders for Physical AI</b></h3>-
<p><strong>Gijs de Jong, Peter Prettenhofer, Rerun</strong><br />-
<i><span style="font-weight: 400;">October 21, 17:30 | 210BF</span></i></p>-
<p><span style="font-weight: 400;">To help developers avoid GPU starvation when training large physical AI models, this session explores how to manage multi-modal temporal datasets, balance decoding and network tradeoffs, and optimize PyTorch dataloaders for complex access patterns. </span></p>-
<h3><b>Building Portable, Composable Local Agents with ExecuTorch</b></h3>-
<p><strong>Mergen Nachin, Digant Desai, Meta</strong><br />-
<i><span style="font-weight: 400;">October 20, 17:45 | LL20CD</span></i></p>-
<p><span style="font-weight: 400;">ExecuTorch serves as a fast, memory-efficient, and hardware-portable runtime substrate that enables efficient, composable, and private local agent experiences across diverse models, devices, and open ecosystem standards. </span></p>-
<h3><b>From PyTorch to the Edge: Agentic Synthesis of Inference Runtimes for Heterogeneous Hardware</b></h3>-
<p><strong>Thomas Cottenier, Arm</strong><br />-
<i><span style="font-weight: 400;">October 21, 17:30 | LL20AB</span></i></p>-
<p><span style="font-weight: 400;">Instead of one hand-tuned general-purpose runtime, an agentic harness synthesizes a bespoke runtime per target. Validated against llama.cpp and MLX on Apple silicon, then pointed at hardware with no baseline at all.</span></p>-
<h3><b>Scaling Audio AI Infrastructure: Building Voice-Native AI</b></h3>-
<p><strong>Mu Li, Huapeng Zhou, Lindsey Allen, Boson AI</strong><br />-
<i><span style="font-weight: 400;">October 21, 12:20 | LL20AB</span></i></p>-
<p><span style="font-weight: 400;">To advance real-time voice-native AI, this system-level discussion explores end-to-end PyTorch post-training and latency-constrained serving optimizations, including extending SGLang for audio workloads. </span></p>-
<h3><b>Sponsored: Hardware-Aware AI: Building Agentic Systems from Cloud to Edge</b></h3>-
<p><strong>Kavya Sri Chennoju, Arm</strong><br />-
<span style="font-weight: 400;">October 20</span><i><span style="font-weight: 400;">, 12:20 | LL20CD</span></i></p>-
<p><span style="font-weight: 400;">This session explores building hardware-aware physical AI applications by combining PyTorch, ExecuTorch, vLLM, and Arm Device Connect in an end-to-end cloud-to-edge workflow that enables models to reason, retrieve live sensor data, and coordinate physical devices w…-
<h2><b>Governance: A practical approach to accountability</b></h2>-
<p><span style="font-weight: 400;">The governance track takes a clear-eyed and practical approach to accountability.</span></p>-
<h3><b>Who Owns Production When the Agent Does the Fixing? Governance in AI-Assisted SRE</b></h3>-
<p><strong>Prakshal Doshi, Aditi Mewada, Apple</strong><br />-
<i><span style="font-weight: 400;">October 20, 11:10 | LL21ABC</span></i></p>-
<p><span style="font-weight: 400;">When an agent misconfigures a service at 3:00 AM and takes down production, who is accountable? This talk offers a practical mental model for what agents should fix, what they should flag, and what should never leave a human’s hands.</span></p>-
<h3><b>Compute, Latency, and Safety: Architecting Stateful Guardrails for Multi-Agent Workflows</b></h3>-
<p><strong>Purva Chiniya, Amazon</strong><br />-
<i><span style="font-weight: 400;">October 20, 11:45 | LL21ABC</span></i></p>-
<p><span style="font-weight: 400;">To secure enterprise multi-agent architectures without destroying throughput, this session details how to build continuous red-teaming pipelines and deploy low-latency Small Language Models as deterministic guardrails to catch indirect orchestration exploits and pr…-
<h3><b>Guardrails at the Infrastructure Layer (Birds of a Feather)</b></h3>-
<p><strong>Sai Charan Teja Gopaluni, Aaresh Sharma, AWS</strong><br />-
<i><span style="font-weight: 400;">October 20, 15:50 | Community Expo</span></i></p>-
<p><span style="font-weight: 400;">This Birds of a Feather session explores infrastructure-level approaches to governing agentic AI in production through container sandboxing, GPU resource boundaries, and token-level observability. </span></p>-
<h3><b>Closing the Confidence Gap: Making AI Output Trustworthy in Enterprise Systems</b></h3>-
<p><strong>Ravi Teja Prabhala Venkata, Capital One</strong><br />-
<i><span style="font-weight: 400;">October 20, 12:20 | LL21ABC</span></i></p>-
<p><span style="font-weight: 400;">To close the AI confidence gap, this session presents a four-layer validation architecture that treats model trustworthiness as an engineering discipline through schema-based type coercion, confidence calibration, contract validation, and continuous observability. …-
<h2><b>Explore the full program</b></h2>-
<p><span style="font-weight: 400;">Join us for two days in San Jose, October 20–21, 2026. Experience agentic RL frameworks being open sourced on stage, kernel agents with published benchmarks, robot policies running on microcontrollers, and maintainers debating open source agent contributions in pub…-
<p><span style="font-weight: 400;">Day one closes with the Flare Party and poster presentations. Day two closes with the AI Community Bash, uniting the PyTorch, AGNTCon, and MCPCon communities alongside a live set from </span><b>De La Soul</b><span style="font-weight: 400;">. Space for the concert i…-
<p><span style="font-weight: 400;">Registration for PyTorch Conference North America 2026 is open now. Visit the official PyTorch Foundation conference page to secure your registration and reserve your hotel.</span></p>-
<p><a href="https://hubs.ly/Q04tDw_W0"><span style="font-weight: 400;">Register for PyTorch Conference North America 2026</span></a></p>-
]]></content:encoded>-
-
-
</item> </channel></rss>-
<!-- plugin=object-cache-pro client=phpredis metric#hits=4251 metric#misses=31 metric#hit-ratio=99.3 metric#bytes=1446047 metric#prefetches=248 metric#store-reads=36 metric#store-writes=8 metric#store-hits=257 metric#store-misses=15 metric#sql-queries=12 metric#ms-total=582.89 metric#ms-cache=22.80 …+
<!-- plugin=object-cache-pro client=phpredis metric#hits=4241 metric#misses=31 metric#hit-ratio=99.3 metric#bytes=1425447 metric#prefetches=248 metric#store-reads=36 metric#store-writes=8 metric#store-hits=257 metric#store-misses=15 metric#sql-queries=12 metric#ms-total=559.66 metric#ms-cache=21.53 …29 lines shown here cut at 300 characters. The raw artifact at this commit is linked above.